Replicate
Run, fine-tune and deploy open-source ML models through a cloud API
Own Replicate? Claim it, free to correct the facts and get the verified badge.
Pricing
Paid
API
Yes
Platforms
What is Replicate?
Replicate is a cloud platform for running machine learning models through an API. It hosts thousands of community-published models for images, video, speech, music and language, plus some proprietary models. Developers can call any of them with a few lines of code using the Python, JavaScript or HTTP clients.
Beyond running public models, you can fine-tune models such as SDXL on your own data, or package your own model with Cog, Replicate's open-source packaging tool, and deploy it on Replicate's infrastructure. Deployments scale up with traffic and down to zero when idle, and logging and monitoring are included.
Billing is pay-as-you-go. Most models are charged per second of hardware time, and some are charged per image, per token or per second of output video. Replicate fits developers and product teams that want to add AI features without running their own GPU servers. Enterprise plans are also offered.
Key features
- Thousands of public models for image, video, audio and text
- Python, JavaScript and HTTP API clients
- Fine-tune models such as SDXL on your own data
- Deploy custom models packaged with open-source Cog
- Autoscaling that scales to zero when idle
- Per-second billing across CPU and Nvidia GPUs
- Logging and monitoring of predictions
Use cases
Who it's for
Adding image generation to an app
Call models like FLUX through the API and pay per output image, with no GPU servers to run.
Deploying a custom model
Package a model with Cog and deploy it on Replicate, which scales it with traffic.
Personalized fine-tunes
Train an image model on photos of a person, product or style, then generate matching images.
Prototyping with open-source LLMs
Try language models such as DeepSeek-R1 with per-token pricing before committing to infrastructure.
Replicate pricing
From the vendor's site, checked Oct 26, 2026; we re-check prices monthly. Prices change — confirm on their pricing page.
Pay-as-you-go (hardware)
usage-based
From $0.000025/sec (CPU Small, $0.09/hr); Nvidia T4 $0.81/hr; H100 $5.49/hr
- Billed per second of run time
- Private models are billed for all time online
- Fast-booting fine-tunes are billed only for active processing time
Pay-as-you-go (per-model)
usage-based
e.g. flux-dev $0.025/image, flux-1.1-pro $0.04/image
- Some models are billed per image, per token or per second of output video
Enterprise
contract
Custom
- Dedicated account manager and priority support
- Higher GPU limits and performance SLAs
- Volume discounts
How Replicate compares
- Replicate has no free plan, while 3 of the 4 alternatives below have one.
- It offers an API, which only 2 of the others do.
| Tool | Starts at | Free plan | Free trial | API | Open source | |
|---|---|---|---|---|---|---|
| R | — | — | — | — | ||
| S | — | — | Replicate vs Stable Diffusion | |||
P Pipedream | $99/mo | — | — | Replicate vs Pipedream | ||
| M | $8/mo | — | — | — | — | Replicate vs Midjourney |
A Adobe Firefly | — | — | — | Replicate vs Adobe Firefly |
Compare 7 capabilities
| Capability | Replicate | Pipedream |
|---|---|---|
| MCP server | ||
| Webhooks | — | |
| Official SDKs | ||
| Bring your own key | — | |
| Custom code steps | — | |
| SSO & RBAC | — | |
| SOC 2 / HIPAA | — |
Capabilities as stated on each vendor's own site; — means the site doesn't say. The closest alternatives in our directory by category. “Starts at” is the cheapest paid plan per month, as each vendor lists it.
Replicate data and privacy
Trains AI on your content
Not stated
Data retention
Retained as long as necessary to provide Services; backups kept limited period
“We generally retain customer personal information for as long as necessary to provide our Services.”
Where data is stored
Not stated
Security and compliance
None stated
From Replicate's own privacy and security pages , checked Oct 28, 2026. We only show what the vendor states in writing; “not stated” means the pages don't say.
Who's behind Replicate
- Company
- Replicate
- Part of
- Cloudflare
“Find our available positions on the Cloudflare careers page .”
From Replicate's own about, terms and privacy pages and Wikidata, checked Oct 28, 2026.
Replicate release activity
6 changelog entries since Dec 2025, latest Apr 21, 2026. Changelog
Pros and cons
Strengths
- Pay only for the compute time or outputs you use
- Wide range of hardware, from CPU to 8x H100
- Cog packaging tool is open source
- Model pages show cost estimates
Limitations
- No free plan listed on the pricing page
- Private models are billed for idle and setup time
- Larger multi-GPU capacity requires committed spend contracts
Replicate FAQ
How much does Replicate cost?
Pricing is pay-as-you-go. Hardware ranges from $0.000025/sec for a small CPU to $0.0122/sec for 8x H100, and some models are priced per image, per token or per second of video.
Is there a subscription?
No subscription tiers are listed. You pay for usage, and enterprise contracts with volume discounts are available.
How do I use it?
Through the API, using the Python or JavaScript client libraries or plain HTTP.
Can I deploy my own model?
Yes. You package it with Cog, an open-source tool, and deploy it on Replicate's infrastructure.
How are private models billed?
For all the time they are online, including setup and idle time. Fast-booting fine-tunes are the exception: they are billed only for active processing.
User reviews
No reviews yet. Used Replicate? Be the first.
Sign in to rate this tool and share how you use it.
Replicate
Paid
Tools similar to Replicate
Open image generation models you can self-host or call through an API
Managed auth and 10,000+ API tools for developers building AI agents
Text-to-image and video generation for artists, designers and creators